We try to capture in the microsimulation model both these behavioural processes and the rates at which various behaviours, events, and changes of status occur.
The basic model is a conventional woman-based microsimulation model of human reproduction and mortality. It differs from earlier models in terms of the heterogeneity of various input parameters and distributions, of the greater complexity and realism of the physiological processes that are taken into account, of its taking into account selected attributes (such as age) of sexual partners and husbands, and in terms of the ease with which the user can incorporate variations in such factors as nuptiality, breastfeeding practices, and infant and child mortality. The basic model of human
reproduction is “female-dominant” in the sense that men are taken into account only implicitly: sterility, for example, is actually an attribute of couples rather than women although treating it as an attribute only of women is both conventional and does not invalidate the model.
This model was subsequently extended by including the sexual transmission of disease, including HIV, so that coitus carries not just a risk of pregnancy to (fecund) women but a risk to all women (whether fecund or not) of disease transmission and acquisition, and also a risk to all men. A novel and useful feature of the model is that simulated disease outcomes are therefore consistent with simulated fertility outcomes.
The simulation takes individuals through their lives, decisions being constantly made, conditional on their past experience and current status, about what will happen next. Women engage in premarital affairs, marry, divorce and remarry according to probability distributions derived from the MDICP data. Both they and their husbands may seroconvert or die, either from AIDS or, since the model incorporates background mortality, from some other cause. Some women have affairs while they are married, and some have affairs after one marriage has ended and before the next one begins. At the same time, women conceive, bear, breastfeed, wean, and lose children, although related tabulations are not shown here because this aspect of their lives is not the focus of the current investigation. The present application of the model is merely one of many. More detail about the workings of the model is presented elsewhere, only the essential features of the present application being presented here.
We try to summarize the input parameters and distributions that govern the simulation of women’s nuptial and sexual behaviour. The timing of their first marriage is simulated according to a region-specific Coale-McNeil marriage model fitted to the MDICP2 data, while that of their husband is derived from the reported age difference between spouses. After the divorce or are widowed, women remarry according to the remarriage functions. In the absence of strong evidence to the contrary and for the sake of simplicity we assume that second and subsequent marriages dissolve at the same rate as first marriages, and that remarriage rates are also undifferentiated by marriage order.
Ninety per cent of women have the propensity to have sex before marriage or after a marriage has ended; ten per cent do not, and will not engage in such affairs. A woman with such a propensity, which is a lifetime characteristic that is assigned just once, has a small annual probability of conducting an affair if she is not currently married, rising to a maximum at age 17 and remaining fixed thereafter. Note that it is not inevitable that a woman with such a propensity will actually embark on such an affair: we might expect, for example, that 63 (0.90 x [1.00-0.30]) per cent of unmarried 17-year-olds will not.
The assigned propensity to engage in extramarital relationships is considerably lower than that of having an affair while unmarried, but extramarital affairs are simulated in a similar manner to non-marital affairs except that the schedule of annual probabilities of entering such a relationship is duration-specific rather than age-specific. In each case, coital frequency is set at half the marital level.
Men’s median age at premarital sexual debut is set at 17 years. We simulate their sexual activity while unmarried differently from the way we simulate the sexual activity of unmarried women because it is portrayed in the journals more as a series of opportunistic encounters than of affairs as such, and also because there is no clear analogue to the bar girl. We posit separate propensities to have sex with casual partners, or “peers”, and with bar girls, and give men with such a propensity a monthly probability, rising from age 15 to age 20, of at least one sexual encounter during that month, and a randomly assigned coital frequency within certain pre-defined limits.
The affairs of married men are simulated in a similar manner to those of married women, although naturally the values of parameters are rather different. Once married, men are less likely to have affairs and less likely to visit bar girls, and their monthly probability of visiting bar girls declines as they age.
For simplicity’s sake, and because condom use is low in rural Malawi, the models assume no use of condoms and no use of contraception more generally.
To introduce infection into the simulated population we posit the existence of a group of women external to the simulated population among whom certain proportions are infectious with various diseases, not just HIV: three discharge (non-ulcerative) STDs (gonorrhoea, chlamydia and bacterial vaginosis), and three ulcerative STDs (syphilis, chancroid and herpes). These women — in the real world, casual partners and bar girls — form the “reservoir of infection” from which disease is introduced into the population at large: they infect men, and these men in turn infect other women. These non-HIV STDs differ according to their effects on HIV transmission and acquisition, and exhibit a degree of heterogeneity in terms of their transmission probabilities, their duration of infectiousness, the likelihood (as with herpes) of spontaneous recurrence, and their conferring of immunity to subsequent infection (significant with syphilis but not, for example, with gonorrhoea). A highlight of the table is the tiny transmission probability of HIV both in absolute terms and in comparison with that of the other diseases included in the model; it therefore remains rather small even in the presence of those other diseases.
Suffice it to say here that, although many of the parameters are not well known, the proportion of bar girls set to be HIV positive is drawn directly from Malawian studies.
The properties of the non-HIV diseases we include in our simulations imply that, even when couples are faithful after marriage, it is possible for a disease to travel from one to the other and then back again: a husband can infect his wife with gonorrhoea, for example, and if she remains infectious for long enough she can re-infect him once his initial infection has passed. It is also possible, except for genital herpes and HIV, for an infection to die out from a marriage if spouses are faithful to one another.
